1 Organic Solvent Nanofiltration of Binary Vegetable Oil/Terpene Mixtures: 1 Experiments and Modelling 2 M. H. Abdellah a,c , L. Liu a , C. A. Scholes a , B. D. Freeman b , S. E. Kentish a * 3 a Department of Chemical Engineering, The University of Melbourne 3010, Australia 4 b Department of Chemical Engineering, The University of Texas at Austin, Texas 5 78712, United States 6 c Department of Chemical Engineering, Faculty of Engineering, Alexandria University, 7 Alexandria, 21544, Egypt 8 9
Transcript
1
Organic Solvent Nanofiltration of Binary Vegetable Oil/Terpene Mixtures: 1
Experiments and Modelling 2
M. H. Abdellaha,c, L. Liua, C. A. Scholesa, B. D. Freemanb, S. E. Kentisha* 3
a Department of Chemical Engineering, The University of Melbourne 3010, Australia 4
b Department of Chemical Engineering, The University of Texas at Austin, Texas 5
78712, United States 6
c Department of Chemical Engineering, Faculty of Engineering, Alexandria University, 7
Alexandria, 21544, Egypt 8
9
2
Abstract 10
Bio-derived solvents such as ρ-cymene, d-limonene and α-pinene represent feasible alternatives to n-11
hexane for the extraction of vegetable oils. However, the large-scale utilization of these solvents is 12
still limited mainly owing to their high boiling points and latent heats of vaporization. In this work, the 13
performance of composite polydimethylsiloxane/polyacrylonitrile (PDMS/PAN) organic solvent 14
nanofiltration membranes in the recovery of these solvents from their binary mixtures with canola oil 15
is investigated. The sorption isotherms of the mixtures were first studied using free-standing PDMS 16
films and the multicomponent Flory-Huggins model used to determine the resulting interaction 17
parameters. The partial solvent uptake decreased with increasing oil concentration in the mixture. On 18
the other hand, the partial oil uptake in the solvent mixture was higher than that of the pure oil which 19
was attributed to the swelling effects induced by solvents. The effects of feed concentration (10-30 20
wt. % oil), feed temperature (25-40 °C), transmembrane pressure (5-30 bar), and cross-flow velocity 21
(18-52 cm s-1) on the membrane performance were then studied in a cross-flow membrane setup. 22
Maxwell-Stefan formulations were combined with the ternary Flory-Huggins solubility model to 23
Vegetable oil extraction and purification is an important industry, with an average annual production 29
of more than 380 million metric tonnes [1]. In the primary stages, the crude oil is extracted from 30
oilseeds by means of an organic solvent; typically n-hexane. N-hexane has been used for many years 31
due to its high solubility for oil, availability and low price. Importantly, its low boiling point (69 °C) and 32
heat of vaporisation (30 KJ mol-1) make the recovery by conventional evaporation feasible[2]. 33
However, n-hexane has recently been classified as a CMR (carcinogenic, mutagenic and reprotoxic) 34
solvent and may be banned for industrial use in the future[3–5]. Moreover, the high volatility and low 35
flash point of n-hexane increase the risk of fire and explosion[6]. Therefore, finding environmentally 36
friendly alternative solvents has become imperative. 37
Bio-derived solvents, particularly terpenes represent promising substitutes for n-hexane [7]. These 38
solvents are extracted from sustainable agricultural materials and have been proved to be safer to 39
humans and the environment [8]. These solvents are used widely in perfumes, pharmaceuticals and 40
in the food industry as flavouring agents [9–11]. Several studies have reported the utilization of 41
terpenes for the extraction of vegetable oil and microalgae lipids[12–18]. These studies concluded 42
that terpenes can replace n-hexane as extractive solvents due to their affinity for oil and lipids 43
constituents. However, their high boiling points and heats of vaporization restrict their utilization at 44
an industrial scale [19]. Furthermore, exposure to high temperatures during the recovery of these 45
solvents using conventional evaporation results in the thermal decomposition of heat sensitive 46
components such as natural antioxidants and the formation of harmful peroxides and unsaturated 47
aldehydes [19,20]. Therefore, finding an alternative solvent recovery process with minimal energy 48
consumption has become inevitable. 49
Membrane separation technologies are increasingly used for aqueous separations in the food 50
industry. This success has triggered interest in such membrane technology for organic solutions, an 51
4
approach known as solvent resistant nanofiltration (SRNF), or organic solvent nanofiltration (OSN). 52
SRNF represents a feasible alternative for energy-intensive conventional separation processes such as 53
distillation, evaporation and liquid-liquid extraction. SRNF has advantages over these conventional 54
processes of low energy consumption, mild operating temperatures, ease of installation and operation 55
and low operating cost[21]. It is also considered as an environmentally friendly technology that 56
minimizes harmful emissions as well as carbon dioxide to the atmosphere [1,22,23]. In the vegetable 57
oil industry, SRNF has the potential to replace different stages of crude oil processing such as solvent 58
recovery, degumming, de-acidification, deodorization and decolourisation[22]. 59
The utilisation of membrane technology in the recovery of solvents from vegetable oil mixtures has 60
been reported in some studies [20,24–31]. Kuk et al. [32] investigated the recovery of ethanol from 61
ethanol-cotton oil mixtures using a commercial reverse osmosis membrane. Tres et al.[33] used a 62
polymeric ultrafiltration membrane for the separation of refined soybean-hexane mixtures and 63
observed a permeate flux up to 65 kg m-2 hr-1 and oil rejections up to 30 %. In another study, Tres et 64
al. [34] observed higher rejections and fluxes for soybean oil/n-butane mixtures relative to oil/n-65
hexane mixtures using ceramic membranes (MWCO 5 and 10 kDa), Raman et al.[35] investigated the 66
separation of 20 wt. % soybean oil in hexane using a polymeric nanofiltration membrane in a 67
multistage process which resulted in a net oil recovery of 99 %. In our own prior work, we investigated 68
the permeation of pure terpenes through a polydimethylsiloxane (PDMS) membrane and observed a 69
non-linear increase in permeate flux that was readily explained using the solution-diffusion model. 70
However, this work considered the permeation of terpenes alone and not oil/solvent mixtures[36]. 71
While the solution-diffusion model is effective for single component systems, it ignores coupling 72
effects between penetrants that occur in binary mixtures[37]. For this reason, it fails to predict the 73
negative solute rejection which is observed in some studies[38,39]. The Maxwell-Stefan (M-S) 74
approach overcomes these limitations[40]. These equations were originally developed to describe 75
multicomponent diffusion through low-density gases but have since been extended successfully to 76
5
dense gases, liquids and polymers[40,41]. The approach has been widely used for describing transport 77
in reverse osmosis, pervaporation, gas permeation and organic solvent nanofiltration systems[42–46]. 78
The M-S equations originate from applying an inter-species force balance and can be written as 79
follows[47]. 80
xi
RT∇T,P μi = − ∑
xixj(ui − uj)
Ðij
m
j=1j≠i
(1)
where xi, μi and ui are the mole fraction, chemical potential (J mol-1) and velocity (m s-1) of component 81
i, respectively. Ðij (m2 s-1) are the M-S diffusivities which represent the inverse of the drag coefficient 82
between species i and j. In dealing with a system where a polymer is one component, it is more 83
convenient to express the M-S equations in terms of volume fractions to be consistent with the 84
thermodynamic models that describe species interactions with polymers[48]. Heintz and co-workers 85
[49] adapted the M-S equations to be in terms of species volume fraction (ϕi) as follows: 86
ϕi
RT∇T,P μi = − ∑
ϕiϕj(ui − uj)
Ðij
m
j=1j≠i
(2)
The formula developed by Heintz has been widely used in modelling multicomponent transport in 87
pervaporation[50–53]. However, the formula does not satisfy the Gibbs-Duhem relationship as 88
pointed out by Fornasiero et al.[54]. To overcome the drawback of Eq. 2, Fornasiero et al.[54] 89
reformulated the original M-S equation (Eq. 1) in terms of the molar segment concentration. With the 90
assumption of no volume change of the multicomponent system upon mixing, Fornasiero obtained 91
the following expression: 92
Ci
RT∇T,P μi = − ∑
ϕiϕj(ui − uj)
υDij°
m
j=1j≠i
(3)
6
where Ci is the molar concentration of the segment i (mol cm-3), υ is the segment molar volume (cm3 93
mol-1) and Dij° is the MS diffusivity of the segments comprising species i and j.Several studies have 94
reported the application of Eq. 3 to modelling the transport of water through polymeric materials [55–95
57], diffusion of asphaltene[58–60], and the formation of polymeric nanoparticles[61]. The only 96
limitation of Eq. 3 is that a reference molar volume consistent with the thermodynamic model must 97
be chosen. Ribeiro et al. [42] adapted the original M-S equations in terms of volume fraction to 98
overcome the limitations of Heintz[49] and Fornasiero [54] and the following relation is obtained: 99
ϕi
ViRT∇T,P μi = − ∑ (
Ni ϕj
Dij
−NjϕiVj
DijVi
)
m
j=1j≠i
(4)
where Vi and Ni are the partial molar volume (cm3 mol-1) and the molar flux (mol cm-2 s-1) of component 100
i, respectively. Dij is the modified M-S diffusivity, cm2 s-1. This formulation assumes that the friction 101
factor between species are not symmetric (i.e. Dij ≠ Dji)[42]. While this is generally not an issue for 102
steady-state transport through a membrane (Nm=0), it would make difficult a treatment of transient 103
phenomena with Nm ≠ 0. Ribeiro et al.[42] used Eq. 4 combined with the Flory-Huggins model to 104
develop a set of ordinary differential equations to describe the steady-state transport of CO2/C2H6 105
mixtures in a cross-linked poly(ethylene oxide) membrane. These differential equations were solved 106
by numerical methods. In an attempt to avoid the complexity associated with solving such differential 107
equations, Krishna [44] used a linearized M-S formulation for the transmembrane fluxes. This 108
expression combines the original M-S equations and the Flory-Huggins model and it provides 109
comparable accuracy. 110
In this work, we aim to investigate the feasibility of SRNF for the recovery of terpenes from their binary 111
mixture with canola (rapeseed) oil under different experimental conditions. Free-standing PDMS films 112
are prepared and used to determine the sorption isotherm of terpenes and their binary mixtures. The 113
performance of PDMS/PAN composite membranes for the recovery of terpenes from canola oil 114
7
solutions is then investigated. The linearized Maxwell-Stefan formulation combined with the Flory-115
Huggins model is used to model these results. 116
2. Model development 117
According to the Flory-Huggins model for a ternary system (polymer (m), solvent (s) and oil(o)), the 118
activity of the solvent (as) and the oil (ao) are related to volume fractions assuming concentration 119
independent interaction parameters as follows[48,62]: 120
ln as = ln ϕs + (1 − ϕs) −
ϕoVs
Vo
−ϕmVS
Vm
+ (χosϕo + χsmϕm)(ϕo + ϕm) − χom
VS
Vo
ϕoϕm (5)
ln ao = ln ϕo + (1 − ϕo) −
ϕsVo
Vs
−ϕmVo
Vm
+ (χos
ϕsVo
Vs
+ χomϕm)(ϕs + ϕm) − χsm
Vo
Vs
ϕsϕm (6)
and 121
ϕo + ϕs + ϕm = 1 (7)
where χos is the mutual interaction parameter between the oil and the solvent. 122
For unidimensional binary mixture permeation across a polymeric membrane given that the 123
membrane is a stationary phase (NmV = 0), Eq. 4 can be rewritten as follows: 124
−ϕs
RT
dμs
dz =
NsV ϕo − No
V ϕs
Dso
+Ns
V ϕm
Dsm
(8)
−
ϕo
RT
dμo
dz =
NoV ϕs − Ns
V ϕo
Dos
+No
V ϕm
Dom
(9)
where NsV and No
V are the volumetric flux of the solvent and the oil (L m-2 hr-1), respectively. 125
8
Starting with Eqs. 8 and 9 Krishna [44] developed a linearized solution for the volumetric fluxes of a 126
binary mixture which can be adapted for the solvent and oil fluxes as follows: 127
NsV =
1
A L[B1(Γ11Δϕs + Γ12Δϕo) + B2(Γ21Δϕs + Γ22Δϕo)] (10)
No
V =1
Α ϕmL[B3(Γ11Δϕs + Γ12Δϕo) + B4(Γ21Δϕs + Γ22Δϕo)]
(11)
where B1 =ϕs
Dos+
ϕm
Dom, B2 =
ϕs
Dso, B3 =
ϕo
Dos, B4 =
ϕo
Dso+
ϕm
Dsm, Γ11 = ϕs
∂ ln as
∂ϕs, Γ12 = ϕs
∂ ln as
∂ϕo, Γ21 =128
ϕo∂ ln ao
∂ϕs, Γ22 = ϕo
∂ ln ao
∂ϕo, A = ϕm (
ϕs
DosDsm+
ϕo
DsoDom+
ϕm
DsmDom),
Dso
Vo=
Dos
Vs, L is the membrane 129
thickness, Δϕs and Δϕo are the volume fraction difference across the membrane for the solvent and 130
the oil, respectively. 131
The thermodynamic factor Γ can be evaluated by the differentiation of the Flory-Huggins model (Eqs. 132
5 -7). Mulder et al.[62] provided explicit expressions for the derivatives ∂ ln ai ∂ϕj⁄ . The arithmetic 133
mean of the volume fractions of the penetrant inside the membrane at the feed and permeate 134
interface ϕi
F+ϕiP
2 is used to calculate (B) and (Γ) values as proposed by Krishna[44]. 135
Dso and Dos represent mutual solvent-oil interactions and are termed exchange coefficients[63,64]. 136
All numerical simulations were conducted using MATLAB® R2014b. 137
3. Materials and methods 138
3.1. Materials 139
Polyacrylonitrile (PAN) membranes (MWCO=20 kDa) were purchased from AMI, USA. A PDMS kit 140
consisting of a pre-polymer (KE106) and a crosslinker (CAT-RG) was supplied by Shin-Etsu Chemicals 141
(Japan). Food grade ρ-cymene, d-limonene and α-pinene were obtained from Sigma-Aldrich. A refined 142
canola oil (Pure Vita - Australia) was used for the preparation of solvent-oil mixtures. The oil consists 143
9
of a mixture of triglycerides (6 wt. % unsaturated C16 and C18, 67 wt. % monounsaturated C18 and 26 144
wt. % polyunsaturated C18) with an average molecular weight of 877 g mol-1. N-hexane and n-heptane 145
were purchased from ChemSupply. All chemicals were used as received without further purification. 146
3.2. Preparation and characterisation of PDMS/PAN membranes 147
Composite PDMS/PAN membranes were prepared by the solution casting method and described in 148
detail in our earlier work [36]. A 500-μm polymer film of a pre-crosslinked 7 % PDMS solution was 149
cast on a wet PAN ultrafiltration support. The polymer solution was prepared in n-heptane with a 150
polymer to crosslinker ratio of 2:1 (w/w). The membrane was thermally cured at 70 °C for 2 hrs, then 151
at 120 °C for another 2 hrs. The final thickness of the PDMS layer was 23 ± 3 μm[36]. For the sorption 152
study, free-standing PDMS films (average dry thickness 1.97 ± 0.04 mm) were prepared from 50 wt. 153
% polymer solution in n-hexane with the same polymer to crosslinker mass ratio. The films were 154
thermally cured at the same conditions used for the preparation of the composite membranes. The 155
average density of PDMS films at 25 °C was estimated from mass measurements in air and distilled 156
water as 1.022 ± 0.005 g cm-3. 157
3.3. Sorption and swelling study 158
The sorption isotherm was studied by the gravimetric method at atmospheric pressure at 25 °C and 159
40 °C. Pre-weighed PDMS films were soaked in oil-solvent mixtures of concentrations ranging from 0 160
to 100 wt. % oil and kept in an oven at the specified temperature. After a specified interval of time (24 161
hrs), the films were taken out, wiped gently with a tissue paper and reweighed. Each measurement 162
was conducted within 30 seconds at most to minimize the experimental error resulting from solvent 163
losses. The process was continued until a constant weight of the films was observed. The total degree 164
of swelling (% SD) was calculated from the equilibrium masses as follows: 165
10
%SD =ma − mb
mb
× 100 (12)
where ma and mb are the equilibrium and initial masses (g) of the PDMS film, respectively. 166
At the end of the sorption study, the partial uptake of both the oil and the solvent was obtained by a 167
complete evaporation of the solvent sorbed by the films. This was accomplished by placing the films 168
under vacuum at 50 °C for a sufficient time (4 days) until a stable mass of the films was observed. The 169
difference between the mass after drying (mad) and the initial mass of the film (mb) gives the partial 170
oil uptake (mo) as follows: 171
mo = mad − mb (13)
Accordingly, the partial solvent uptake (ms) is calculated as follows: 172
ms = ma − mad (14)
3.4. Nanofiltration experiments 173
Nanofiltration of solvent/oil mixtures was conducted in a cross-flow cell (Sterlitech-CF042) as 174
described in our earlier work [36]. The feed mixture was circulated by a high-pressure diaphragm 175
pump (Hydra cell P200). A needle valve was mounted on the retentate line and used to control the 176
pressure acting on the membrane which was measured by a pressure gauge (Floyd). To maintain a 177
constant feed temperature, the feed line was immersed in a temperature-controlled water bath. A 178
PTFE high foulant (65 mils, Sterlitech) feed side spacer was placed on the membrane feed side to 179
minimise the concentration polarisation effects. A fresh membrane was used for each of the solvent-180
oil mixtures. Prior to the experiment, this membrane was conditioned by permeating n-hexane at 30 181
bar for 3 hours followed by the pure terpene solvent for 3 hrs. The binary oil/solvent mixture was then 182
filtered for another 3 hrs at the same pressure. The retentate was circulated back to the feed tank. 183
11
The permeate was collected and weighed on a digital balance (Ohaus) connected to a PC to record the 184
mass measurements at intervals of 5 mins using the TWedge data acquisition software. In order to 185
maintain consistent feed concentration, the permeate collected was returned back to the feed tank 186
after collecting 100 ml. 187
The membrane performance was evaluated in terms of the permeate flux and the oil rejection. The 188
permeate volumetric flux (NPv) was calculated as follows, 189
NP
V =∆mP
ρP∆tA (L m-2 hr-1) (15 )
where ∆mP is the mass (kg) of the permeate collected at time ∆t (hr), A is the membrane active area 190
(42×10-4 m2) and ρP(kg L-1) is the permeate density. The oil retention in the permeate solution was 191
determined by the solvent evaporation method [26,27,34,65]. A 3-ml sample was weighed and placed 192
in a vacuum oven at 120 °C for 4 hours to remove the residual solvent. Afterwards, the sample was 193
left to cool down at ambient conditions in a desiccator, then reweighed. The apparent oil rejection 194
(% Ra) was calculated from Eq. 16a, while the true rejection (%Rt) was calculated from Eq. 16b: 195
% Ra =cB − cP
cB
× 100 (16a)
% Rt =cF − cP
cF
× 100 (16b)
where cB , cF and cP are the oil concentrations (w/w) in the bulk feed, at the feed side of the membrane 196
interface and in the permeate respectively. 197
12
4. Results and discussion 198
4.1. Sorption and swelling 199
The average uptakes of pure pinene, limonene, cymene and canola oil in PDMS films at both 25 °C and 200
40 oC were determined in our previous paper as 3.13, 2.77, 2.58 g/g dry membrane [36]. In the present 201
case, we determined that the uptake of pure canola oil was three orders of magnitude lower at 3.90 202
×10-3 g/g dry membrane. Stafie et al.[31] reported similar behaviour and attributed this to the large 203
molecular size of the oil compared with the solvent. The partial uptakes of the solvents/canola oil from 204
their binary mixtures in PDMS films are depicted in Fig.1. 205
13
Figure 1. Partial uptakes of A) pinene/ canola oil, B) limonene/canola oil and C) cymene/canola oil from their binary 206 mixtures in PDMS films. Symbols represent experimental data while lines are the predictions from the Flory-Huggins 207
model. 208
In the entire range of concentrations, the oil sorption levels were significantly lower than those of the 209
solvents (note the different y-axis scales in Fig. 1). However, the partial oil uptake in oil/solvent 210
mixtures was higher than that of the pure oil, reaching a maximum at 70 wt. % solvent. With the 211
addition of solvent to the oil the degree of polymer swelling increases which in turn enhances the 212
14
sorption levels of oil. Similar behaviour has been observed during the sorption of ethanol/water and 213
propanol/water mixtures in PDMS[66–68]. 214
The ternary Flory-Huggins model was used to predict these sorption isotherms (Fig. 1). The binary 215
Flory-Huggins interaction parameters for the pure solvents, as determined from our prior work were 216
used directly, and the solvent/oil interaction parameters then determined by minimising the error 217
between predicted and experimental data (Table 1) [48,66,69]. All interaction parameters used here 218
were assumed concentration independent. The liquid phase oil/solvent solutions are assumed to 219
behave ideally[31][70] so that the activity coefficient for the liquid mixture was assumed unity. 220
Table 1. Molar volumes and Flory Huggins interaction parameters for the solvents and the canola oil 221
Molar volume,
Vi (cm3 mol-1)
Interaction
parameter, χim
Interaction
parameter, χso
α-pinene 161 0.58 ± 0.01 [36] 0.46 ± 0.02
d-limonene 163 0.59 ± 0.01 [36] 0.38 ± 0.02
p-cymene 158 0.60 ± 0.01 [36] 0.32 ± 0.01
Canola oil 960 4.51 ± 0.03 -----
222
4.2. Nanofiltration experiments 223
4.2.1. Concentration polarisation 224
Concentration polarisation results from the build-up of solute molecules in the boundary layer at the 225
membrane interface[71]. This layer of solute introduces an additional mass transfer resistance which 226
in turn reduce the permeate flux along the operation. Moreover, it contributes into the deterioration 227
15
of the membrane selectivity due to the increased solute concentration at the membrane surface, thus 228
increasing the solute concentration driving force across the membrane. The effects of concentration 229
polarisation can be minimised by operating at high cross-flow velocities and utilisation of a feed spacer 230
[71]. Although several studies have reported the existence of concentration polarisation effects in 231
organic solvent nanofiltration[27,65,72–74], only the work done by Peeva et al. reported a detailed 232
investigation of concentration polarisation effects [75]. 233
The effect of feed cross-flow velocities on the permeate flux was studied in the range of 18-52 cm s-1 234
using hexane as the solvent. An increase in the permeate flux, particularly at higher pressures and 235
feed concentration was observed with increasing the cross-flow velocity of the feed solution 236
(Supplementary Information, Fig. S1). The mass transfer coefficients of the oil within the mass 237
transfer boundary layer were determined using the approach developed by Sutzkover et al.[76]. 238
k =
NPV
ln {∆P
πF − πP × [1 −
NPV
NsV]}
×1
36000 (17)
where k is the oil mass transfer coefficient (cm s-1), NPV and Ns
V are the permeate and the pure solvent 239
fluxes (L m-2 hr-1), respectively, and πF and πP are the osmotic pressures of the feed and permeate 240
solutions, respectively. The osmotic pressures were calculated using the Van’t Hoff equation as 241
follows: 242
π = RTC (18)
where R is the universal gas constant, T is the absolute temperature and C is the molar concentration 243
of the oil. 244
16
Using Eq. 17, the mass transfer coefficients of the oil at various experimental conditions were 245
estimated graphically from the straight-line fit of NPV versus the ln {
∆P
πF−πP × [1 −
NPV
NsV]} as shown in Fig. 246
S2 (Supplementary information). 247
The mass transfer coefficients were then used to develop the mass transfer correlation of the system 248
as (see Supplementary Information): 249
Sh = 0.569 Re0.56Sc0.30 (
dh
L)
0.5
(19)
where Sh is the Sherwood number(kdh
𝒟o), Re is the Reynolds number(
ρvdh
μ), Sc is the Schmidt 250
number(μ
ρ𝒟o), dh is the hydraulic diameter (cm) of the flow channel, L is the length of the flow channel 251
(cm), ρ is the feed solution density (g cm-3), μ is the dynamic viscosity of the feed solution (cP), v is the 252
crossflow velocity of the feed solution (cm s-1) and 𝒟o is the diffusivity of the oil in the solution (cm2 s-253
1). 254
4.2.2. Binary mixture filtration 255
The permeate flux of 10 wt. % and 30 wt. % oil solutions in the different solvents as a function of the 256
net driving force (transmembrane pressure minus the osmotic pressure difference ∆P − (πF − πP)). 257
are provided in Fig. 2 and 3. As with our earlier work, the flux change with pressure is clearly non-258
linear reflecting the change in solvent concentration within the membrane as pressure increases. 259
Consistent with our prior work, oil/limonene solutions showed the highest permeate flux followed by 260
oil/cymene, then oil/pinene. Increasing the oil content in the feed solution resulted in a significant 261
decline in the permeate flux reflecting the greater viscosity of the solution and the lower degree of 262
swelling (Supplementary Information, Table S3). The nonlinear flux-pressure behaviour observed in 263
our previous work is also less evident with increasing oil content in the feed solution, reflecting 264
reduced swelling in these systems. Increasing the temperature resulted in increased flux due to the 265
17
lower viscosity. These results are consistent with other workers[19,27,31,35] who have used similar 266
silicone based polymers. 267
Figure 2. Permeate flux and true oil rejection (% Rt) of 10 wt. % oil solutions as a function the pressure driving force at A) 25 268 °C and B) 40 °C. The apparent oil rejection (based on the bulk oil concentrations) is provided in Fig. S5. 269
270
Figure 3. Permeate flux and true oil rejection (% Rt) of 30 wt. % oil solutions as a function the pressure driving force at A) 25 271 °C and B) 40 °C. The apparent oil rejection (based on the bulk oil concentrations) is provided in Fig. S6. 272
273
18
The apparent oil rejection (Eq. 16a) is provided in Fig. S5 and S6 in the Supplementary Information as 274
a function of the driving force for the different oil-solvent solutions. The true oil rejection (Eq. 16b) is 275
presented in Fig. 2 and 3. This rejection is calculated based on the oil concentration at the feed side 276
of the membrane interface(cF), that is, after taking account of concentration polarisation. This oil 277
concentration was calculated from Eq. 20 using the oil mass transfer coefficients determined using 278
Eq. 19 and presented in Table. S4 in the Supplementary Information. 279
cF − cP
cB − cP= exp (
NPV
k) (20)
An increase in the oil rejection (%) was observed with increasing pressure, reflecting the increase in 280
the solvent flux through the membrane with increasing pressure, while the oil flux remains relatively 281
unchanged. Importantly, the rejection exceeds 80 % for pressures beyond 20 Bar which indicates that 282
SRNF is a viable approach for solvent removal up to 30 wt. % oil. No significant difference in oil 283
rejection between the different oil-solvent mixtures can be observed. Increasing the feed temperature 284
resulted in a slight deterioration in the membrane selectivity. This may result from the lower viscosity 285
of the oil at these temperatures, or the greater free volume inside the membrane. A decrease in oil 286
rejection was also observed with increasing feed oil concentration. 287
4.2.3. Maxwell-Stefan Model simulations 288
The linearized M-S formulations developed by Krishna were then used to simulate the steady-state 289
mass transfer of the oil/solvent binary mixture across the PDMS/PAN membrane at the different 290
experimental conditions. The volume fractions of the oil and solvent at the upstream and downstream 291
faces were obtained by solving the ternary Flory-Huggins model (Eqs. 5-7) using the interaction 292
parameters χsm, χom, and χso presented in Table 1 and after accounting for concentration polarisation 293
on the upstream side. 294
In initial simulations, the solvent diffusivity in the membrane (Dsm) was taken as an adjustable 295
parameter and allowed to vary at each pressure condition. The exchange coefficients Dom and Dos did 296
19
not vary with pressure, but were allowed to change with temperature and feed oil concentration. This 297
approach gave an excellent fit to all data sets as shown in Fig 4. 298
Figure 4. Flux of pinene and oil obtained from the initial set of simulations where the diffusion coefficient was allowed to 299
vary with pressure A) 10 wt. % oil with 𝐷𝑜𝑚 = 1.4 ± 0.5 𝑥 10-6 cm2 s-1 and 𝐷𝑜𝑠 = 3.5 ± 0.4 𝑥 10-7 cm2 s-1 and B) 30 wt. % oil 300 with 𝐷𝑜𝑚 = 0.94 ± 0.4 𝑥 10-6 cm2 s-1 and 𝐷𝑜𝑠 = 2.2 ± 0.7 𝑥 10-7 cm2 s-1. Symbols represent experimental data where lines 301
represent simulation results. 302
303
These preliminary simulations revealed a clear dependency of the solvent diffusivity (Dsm) and the 304
average volume fraction of solvent in the membrane (Fig. 5). Such concentration-dependent diffusion 305
coefficients are commonly observed in the literature and relate to the degree of membrane swelling. 306
They are often described by an exponential relationship as follows [42,77–81]: 307
Dsm = Dsm0 exp(βs ϕs) (21)
where Dsm0 is the penetrant diffusivity in the limit ϕs → 0, βs is an empirical constant reflecting the 308
effect of penetrant concentration on the membrane swelling. As shown in Fig. 5, such a relationship 309
would appear valid here across all oil concentrations (0, 10 and 30 wt. %), although there is clearly 310
some variability related to experimental error. 311
20
312
Figure 5. Pinene diffusivity (Dsm) for 0, 10 and 30 wt. % oil solutions at 25 °C through PDMS/PAN composite 313 membrane as a function of solvent volume fraction in the membrane, determined from the initial set of 314
simulations. 315
316
21
A second set of simulations was thus carried out where (Dsm) was determined from Eq.21. It would 317
have been preferable to determine the Dom exchange coefficient from filtration experiments using 318
pure oil. However, the flux of oil in such experiments was below the level of detection, as the oil 319
solubility in the absence of solvents is extremely low (Table 1). As a result, Dom was set as a fitted 320
parameter. It is worthwhile noting that if the ‘friction’ between oil and solvents was ignored (i.e.) 321
Dso>> Dsm, Dos>>Dom), negative fluxes were obtained, highlighting the importance of ‘friction’ on the 322
oil transport. A choice of Dom
Dos=10 for all experimental conditions provided robust simulation results 323
and it further reduced the number of fitted parameters. The resulting model parameters are given in 324
Table 4 while typical results showing the fit to the experimental data are presented in Figs. 6-8. As it 325
can be seen, both experimental and predicted flux data show a similar trend with deviations due to 326
experimental error. The trend of constant oil flux with pressure occurs because this flux relates mostly 327
to the concentration gradient across the membrane, which does not vary with pressure. On the other 328
hand, the solvent fluxes show a non-linear increase with increasing pressure driving force which can 329
be explained by the membrane swelling as reported by others[36,82]. 330
331
22
Table 4. Fitting parameters of Eqs. 15 and 16 for permeation of oil/solvent binary mixtures through PDMS/PAN membrane. 332
The simulations were conducted for 0, 10 and 30 wt. % oil. 𝐷𝑜𝑠 = 𝐷𝑜𝑚/10. 333
T (°C) Dsm
0 (cm2 s-1) × 106 βs Wt. % oil Dom(cm2 s-1) × 106
Pinene 25 0.9 ± 0.3 3.1 ± 0.6 10 3.1 ± 0.5
30 1.2 ± 0.7
40 1.7 ± 0.7 2.5 ± 0.6 10 4.0 ± 1.0
30 2.3 ± 1.4
Cymene 25 0.8 ± 0.5 4.0 ± 1.0 10 3.8 ± 1.8
30 2.5 ± 0.4
40 2.1 ± 0.4 2.6 ± 0.3 10 6.0 ± 3.0
30 3.6 ± 1.3
Limonene 25 1.5 ± 0.6 3.2 ± 0.6 10 6.0 ± 2.0
30 2.6 ± 0.8
40 3.2 ± 0.6 2.3 ± 0.3 10 7.6 ± 4.7
30 5.7 ± 1.7
334
As shown in Table 4, the pre-exponential term for the solvent/membrane (Dsmo ) is highest for 335
limonene, consistent with the higher flux observed with this solvent. The oil membrane exchange 336
coefficients are also highest for limonene, indicative of greater solvent concentrations when this 337
solvent was used. The oil diffusivity (Dom) and the pre-exponential term in Eq. 21 (Dsmo ) increased with 338
increasing the temperature as expected. The oil diffusivity (Dom) also decreased as the solvent 339
concentration fell (i.e. the oil concentration increased) as expected from swelling effects. 340
23
Figure 6. Flux of pinene and oil through the PDMS/PAN membrane from A) 10 wt. % oil and B) 30 wt. % oil mixtures. 341 Symbols represent experimental data and lines are the simulation results. 342
Figure 7. Flux of cymene and oil through the PDMS/PAN membrane from A) 10 wt. % oil and B) 30 wt. % oil mixtures. 343 Symbols represent experimental data and lines are the simulation results. 344
345
24
Figure 8. Flux of limonene and oil through the PDMS/PAN membrane from A) 10 wt. % oil and B) 30 wt. % oil mixtures. 346 Symbols represent experimental data and lines are the simulation results. 347
5. Conclusions 348
In this work composite PDMS/PAN membranes were prepared and used for the recovering of pinene, 349
limonene and cymene from their binary mixtures with canola oil as a solute. The highest solvent fluxes 350
were observed with oil/limonene mixtures whereas the oil rejections were comparable in the different 351
oil/solvent mixtures and up to 90 % was achieved. The concentration polarisation effects on the 352
system were studied and the mass transfer correlation of the system was developed. The sorption 353
levels of the oil and the solvents were studied and were successfully predicted by the ternary Flory-354
Huggins model with concentration independent interaction parameters. Using linearized MS 355
formulations, the oil and solvent fluxes across the membrane were calculated and were in a good 356
agreement with the experimental observations. The results of this study indicate that SRNF has 357
potential as an alternative to evaporation for the recovery of terpenes from vegetable oil mixtures. 358
Future research will focus on improving the membrane selectivity to meet the industry requirement 359
of more than 95 % oil rejection. This can be achieved by varying the cross-linking density and/or the 360
membrane hydrophobicity by using functionalised PDMS. 361
25
Acknowledgements 362
M. H. Abdellah acknowledges The University of Melbourne for the Melbourne Research Scholarship 363
and Alexandria University for the financial support. L. Liu and S.E. Kentish acknowledge funding 364
support from the Australian Research Council (ARC) Discovery Program (DP150100977). B.D. Freeman 365
gratefully acknowledges support from the Australian-American Fulbright Commission for the award 366
of a U.S. Fulbright Distinguished Chair in Science, Technology, and Innovation sponsored by the 367
Commonwealth Scientific and Industrial Research Organization (CSIRO). 368
References 369
[1] P. Vandezande, L.E.M. Gevers, I.F.J. Vankelecom, Solvent resistant nanofiltration: separating 370 on a molecular level, Chem. Soc. Rev. 37 (2008) 365–405. 371 [2] F.D. Gunstone, J.L. Harwood, A.J. Dijkstra, The Lipid Handbook, 3rd ed., CRC Press, 2007. 372 [3] A. Benazzouz, L. Moity, C. Pierlot, M. Sergent, V. Molinier, J.-M. Aubry, Selection of a 373 Greener Set of Solvents Evenly Spread in the Hansen Space by Space-Filling Design, Ind. Eng. Chem. 374 Res. 52 (2013) 16585–16597. 375 [4] Z. Li, K.H. Smith, G.W. Stevens, The use of environmentally sustainable bio-derived solvents 376 in solvent extraction applications—A review, Chin. J. Chem. Eng. 24 (2016) 215–220. 377 [5] REACH - ECHA, (2016). http://echa.europa.eu/regulations/reach (accessed April 14, 2016). 378 [6] E. K, P. Cuperus F, Solvent Resistant Nanofiltration Membranes in Edible Oil Processing, 379 Membrane Technology. 1999 (1999) 5–8. 380 [7] F. Chemat, M.A. Vian, eds., Alternative Solvents for Natural Products Extraction, Springer 381 Berlin Heidelberg, Berlin, Heidelberg, 2014. 382 [8] F. Kerton, Alternative Solvents for Green Chemistry, Royal Society of Chemistry, Cambridge, 383 2009. 384 [9] L. Caputi, E. Aprea, Use of Terpenoids as Natural Flavouring Compounds in Food Industry, 385 Recent Pat. Food Nutr. Agric. 3 (2011) 9–16. 386 [10] A. Silva Santos, A. Antunes, L. D’Avila, H. Bizzo, L. Souza-Santos, The Use of Essential Oils and 387 Terpenes/Terpeneoids in Cosmetic and Perfumery., Perfum. Flavor. 30 (2005) 50–55. 388 [11] M. Aqil, A. Ahad, Y. Sultana, A. Ali, Status of terpenes as skin penetration enhancers, Drug 389 Discovery Today. 12 (2007) 1061–1067. 390 [12] M.-T. Golmakani, J.A. Mendiola, K. Rezaei, E. Ibáñez, Pressurized limonene as an alternative 391 bio-solvent for the extraction of lipids from marine microorganisms, J. Supercrit. Fluids. 92 (2014) 1–392 7. 393 [13] C. Dejoye Tanzi, M. Abert Vian, F. Chemat, New procedure for extraction of algal lipids from 394 wet biomass: a green clean and scalable process, Bioresour. Technol. 134 (2013) 271–275. 395 [14] C. Dejoye Tanzi, M. Abert Vian, C. Ginies, M. Elmaataoui, F. Chemat, Terpenes as Green 396 Solvents for Extraction of Oil from Microalgae, Molecules. 17 (2012) 8196–8205. 397 [15] A.-G. Sicaire, M. Vian, F. Fine, F. Joffre, P. Carré, S. Tostain, F. Chemat, Alternative bio-based 398 solvents for extraction of fat and oils: solubility prediction, global yield, extraction kinetics, chemical 399 composition and cost of manufacturing, Int. J. Mol. Sci. 16 (2015) 8430–8453. 400
26
[16] P.K. Mamidipally, S.X. Liu, First approach on rice bran oil extraction using limonene, Eur. J. 401 Lipid Sci. Technol. 106 (2004) 122–125. 402 [17] S.X. Liu, P.K. Mamidipally, Quality Comparison of Rice Bran Oil Extracted with d-Limonene 403 and Hexane, Cereal Chem. 82 (2005) 209–215. 404 [18] M. Virot, V. Tomao, C. Ginies, F. Chemat, Total Lipid Extraction of Food Using d-Limonene as 405 an Alternative to n-Hexane, Chroma. 68 (2008) 311–313. 406 [19] S. Arora, S. Manjula, A.G.G. Krishna, R. Subramanian, Membrane processing of crude palm 407 oil, Desalination. 191 (2006) 454–466. 408 [20] R. Subramanian, M. Nakajima, T. Kawakatsu, Processing of vegetable oils using polymeric 409 composite membranes, J. Food Eng. 38 (1998) 41–56. 410 [21] M. Takht Ravanchi, T. Kaghazchi, A. Kargari, Application of membrane separation processes 411 in petrochemical industry: a review, Desalination. 235 (2009) 199–244. 412 [22] M. Cheryan, Membrane technology in the vegetable oil industry, Membrane Technology. 413 2005 (2005) 5–7. 414 [23] P. Marchetti, M.F. Jimenez Solomon, G. Szekely, A.G. Livingston, Molecular Separation with 415 Organic Solvent Nanofiltration: A Critical Review, Chem. Rev. 114 (2014) 10735–10806. 416 [24] J.B. Snape, M. Nakajima, Processing of agricultural fats and oils using membrane technology, 417 J. Food Eng. 30 (1996) 1–41. 418 [25] S.S. Köseoglu, J.T. Lawhon, E.W. Lusas, Membrane processing of crude vegetable oils: Pilot 419 plant scale remoyal of solvent from oil miscellas, J. Am. Oil Chem. Soc. 67 (1990) 315–322. 420 [26] M.V. Tres, J.C. Racoski, R. Nobrega, R.B. Carvalho, J.V. Oliveira, M. Di Luccio, Solvent 421 recovery from soybean oil/n-butane mixtures using a hollow fiber ultrafiltration membrane, Braz. J. 422 Chem. Eng. 31 (2014) 243–249. 423 [27] A.P.B. Ribeiro, J.M.L.N. de Moura, L.A.G. Gonçalves, J.C.C. Petrus, L.A. Viotto, Solvent 424 recovery from soybean oil/hexane miscella by polymeric membranes, J. Membr. Sci. 282 (2006) 425 328–336. 426 [28] J. Chi-Sheng Wu, E.-H. Lee, Ultrafiltration of soybean oil/hexane extract by porous ceramic 427 membranes, J. Membr. Sci. 154 (1999) 251–259. 428 [29] W. Cai, Y. Sun, X. Piao, J. Li, S. Zhu, Solvent Recovery from Soybean Oil/Hexane Miscella by 429 PDMS Composite Membrane, Chin. J. Chem. Eng. 19 (2011) 575–580. 430 [30] S. Darvishmanesh, T. Robberecht, P. Luis, J. Degrève, B. Van der Bruggen, Performance of 431 Nanofiltration Membranes for Solvent Purification in the Oil Industry, J. Am. Oil Chem. Soc. 88 432 (2011) 1255–1261. 433 [31] N. Stafie, D.F. Stamatialis, M. Wessling, Insight into the transport of hexane–solute systems 434 through tailor-made composite membranes, J. Membr. Sci. 228 (2004) 103–116. 435 [32] M.S. Kuk, R.J. Hron, G. Abraham, Reverse osmosis membrane characteristics for partitioning 436 triglyceride-solvent mixtures, J. Am. Oil Chem. Soc. 66 (1989) 1374–1380. 437 [33] M.V. Tres, R. Nobrega, R.B. Carvalho, J.V. Oliveira, M.D. Luccio, Solvent recovery from 438 soybean oil/n-hexane mixtures using hollow fiber membrane, Braz. J. Chem. Eng. 29 (2012) 577–439 584. 440 [34] M.V. Tres, J.C. Racoski, M. Di Luccio, J.V. Oliveira, H. Treichel, D. de Oliveira, M.A. Mazutti, 441 Separation of soybean oil/n-hexane and soybean oil/n-butane mixtures using ceramic membranes, 442 Food Res. Int. 63, Part A (2014) 33–41. 443 [35] L.P. Rama, M. Cheryan, N. Rajagopalan, Solvent recovery and partial deacidification of 444 vegetable oils by membrane technology, Eur. J. Lipid Sci. Technol. 98 (1996) 10–14. 445 [36] M.H. Abdellah, C.A. Scholes, B.D. Freeman, L. Liu, S.E. Kentish, Transport of terpenes through 446 composite PDMS/PAN solvent resistant nanofiltration membranes, Sep. Purif. Technol. 207 (2018) 447 470–476. 448 [37] D.R. Paul, Reformulation of the solution-diffusion theory of reverse osmosis, J. Membr. Sci. 449 241 (2004) 371–386. 450
27
[38] H.K. Lonsdale, U. Merten, M. Tagami, Phenol transport in cellulose acetate membranes, J. 451 Appl. Polym. Sci. 11 (1967) 1807–1820. 452 [39] A.E. Yaroshchuk, Negative rejection of ions in pressure-driven membrane processes, Adv. 453 Colloid Interface Sci. 139 (2008) 150–173. 454 [40] R.B. Bird, W.E. Stewart, E.N. Lightfoot, Transport Phenomena, John Wiley & Sons, 2007. 455 [41] C.F. Curtiss, R.B. Bird, Multicomponent Diffusion, Ind. Eng. Chem. Res. 38 (1999) 2515–2522. 456 [42] C.P. Ribeiro, B.D. Freeman, D.R. Paul, Modeling of multicomponent mass transfer across 457 polymer films using a thermodynamically consistent formulation of the Maxwell–Stefan equations in 458 terms of volume fractions, Polymer. 52 (2011) 3970–3983. 459 [43] F. Fornasiero, J.M. Prausnitz, C.J. Radke, Multicomponent Diffusion in Highly Asymmetric 460 Systems. An Extended Maxwell−Stefan Model for Starkly Different-Sized, Segment-Accessible Chain 461 Molecules, Macromolecules. 38 (2005) 1364–1370. 462 [44] R. Krishna, Describing mixture permeation across polymeric membranes by a combination of 463 Maxwell-Stefan and Flory-Huggins models, Polymer. 103 (2016) 124–131. 464 [45] P. Izák, L. Bartovská, K. Friess, M. Šıpek, P. Uchytil, Comparison of various models for 465 transport of binary mixtures through dense polymer membrane, Polymer. 44 (2003) 2679–2687. 466 [46] P. Marchetti, A.G. Livingston, Predictive membrane transport models for Organic Solvent 467 Nanofiltration: How complex do we need to be?, J. Membr. Sci. 476 (2015) 530–553. 468 [47] R. Krishna, J.A. Wesselingh, The Maxwell-Stefan approach to mass transfer, Chem. Eng. Sci. 469 52 (1997) 861–911. 470 [48] P.J. Flory, Principles of Polymer Chemistry, Cornell University Press, New York, 1953. 471 [49] A. Heintz, W. Stephan, A generalized solution—diffusion model of the pervaporation process 472 through composite membranes Part II. Concentration polarization, coupled diffusion and the 473 influence of the porous support layer, J. Membr. Sci. 89 (1994) 153–169. 474 [50] A. Raisi, A. Aroujalian, T. Kaghazchi, A predictive mass transfer model for aroma compounds 475 recovery by pervaporation, J. Food Eng. 95 (2009) 305–312. 476 [51] S.J. Lue, F.J. Wang, S.-Y. Hsiaw, Pervaporation of benzene/cyclohexane mixtures using ion-477 exchange membrane containing copper ions, J. Membr. Sci. 240 (2004) 149–158. 478 [52] P. Izák, L. Bartovská, K. Friess, M. Šıpek, P. Uchytil, Description of binary liquid mixtures 479 transport through non-porous membrane by modified Maxwell–Stefan equations, J. Membr. Sci. 480 214 (2003) 293–309. 481 [53] V.S. Cunha, M.L.L. Paredes, C.P. Borges, A.C. Habert, R. Nobrega, Removal of aromatics from 482 multicomponent organic mixtures by pervaporation using polyurethane membranes: experimental 483 and modeling, J. Membr. Sci. 206 (2002) 277–290. 484 [54] F. Fornasiero, J.M. Prausnitz, C.J. Radke, Multicomponent Diffusion in Highly Asymmetric 485 Systems. An Extended Maxwell−Stefan Model for Starkly Different-Sized, Segment-Accessible Chain 486 Molecules, Macromolecules. 38 (2005) 1364–1370. 487 [55] F. Fornasiero, F. Krull, J.M. Prausnitz, C.J. Radke, Steady-state diffusion of water through 488 soft-contact-lens materials, Biomater. 26 (2005) 5704–5716. 489 [56] F. Fornasiero, J.M. Prausnitz, C.J. Radke, Post-lens tear-film depletion due to evaporative 490 dehydration of a soft contact lens, J. Membr. Sci. 275 (2006) 229–243. 491 [57] F. Fornasiero, D. Tang, A. Boushehri, J. Prausnitz, C.J. Radke, Water diffusion through 492 hydrogel membranes: A novel evaporation cell free of external mass-transfer resistance, J. Membr. 493 Sci. 320 (2008) 423–430. 494 [58] C. Ferreira, J. Marques, M. Tayakout, I. Guibard, F. Lemos, H. Toulhoat, F. Ramôa Ribeiro, 495 Modeling residue hydrotreating, Chem. Eng. Sci. 65 (2010) 322–329. 496 [59] M. Tayakout, C. Ferreira, D. Espinat, S. Arribas Picon, L. Sorbier, D. Guillaume, I. Guibard, 497 Diffusion of asphaltene molecules through the pore structure of hydroconversion catalysts, Chem. 498 Eng. Sci. 65 (2010) 1571–1583. 499
28
[60] C. Marchal, E. Abdessalem, M. Tayakout-Fayolle, D. Uzio, Asphaltene Diffusion and 500 Adsorption in Modified NiMo Alumina Catalysts Followed by Ultraviolet (UV) Spectroscopy, Energy 501 Fuels. 24 (2010) 4290–4300. 502 [61] M. Hassou, F. Couenne, Y. le Gorrec, M. Tayakout, Modeling and simulation of polymeric 503 nanocapsule formation by emulsion diffusion method, AIChE J. 55 (2009) 2094–2105. 504 [62] M.H.V. Mulder, C.A. Smolders, On the mechanism of separation of ethanol/water mixtures 505 by pervaporation I. Calculations of concentration profiles, J. Membr. Sci. 17 (1984) 289–307. 506 [63] R. Krishna, The Maxwell–Stefan description of mixture diffusion in nanoporous crystalline 507 materials, Microporous Mesoporous Mater. 185 (2014) 30–50. 508 [64] R. Krishna, Diffusion in porous crystalline materials, Chem. Soc. Rev. 41 (2012) 3099–3118. 509 [65] M.V. Tres, S. Mohr, M.L. Corazza, M. Di Luccio, J.V. Oliveira, Separation of n-butane from 510 soybean oil mixtures using membrane processes, J. Membr. Sci. 333 (2009) 141–146. 511 [66] T.-H. Yang, S.J. Lue, Modeling Sorption Behavior for Ethanol/Water Mixtures in a Cross-512 linked Polydimethylsiloxane Membrane Using the Flory-Huggins Equation, Journal of 513 Macromolecular Science, Part B. 52 (2013) 1009–1029. 514 [67] H.J.C. te Hennepe, W.B.F. Boswerger, D. Bargeman, M.H.V. Mulder, C.A. Smolders, Zeolite-515 filled silicone rubber membranes Experimental determination of concentration profiles, J. Membr. 516 Sci. 89 (1994) 185–196. 517 [68] T.-H. Yang, S. Jessie Lue, UNIQUAC and UNIQUAC-HB models for the sorption behavior of 518 ethanol/water mixtures in a cross-linked polydimethylsiloxane membrane, J. Membr. Sci. 415–416 519 (2012) 534–545. 520 [69] M.H.V. Mulder, T. Franken, C.A. Smolders, Preferential sorption versus preferential 521 permeability in pervaporation, J. Membr. Sci. 22 (1985) 155–173. 522 [70] D.F. Stamatialis, N. Stafie, K. Buadu, M. Hempenius, M. Wessling, Observations on the 523 permeation performance of solvent resistant nanofiltration membranes, J. Membr. Sci. 279 (2006) 524 424–433. 525 [71] S.E. Kentish, G. Rice, Demineralization of dairy streams and dairy mineral recovery using 526 nanofiltration, in: K. Hu, J.M. Dickson (Eds.), Membrane Processing for Dairy Ingredient Separation, 527 John Wiley & Sons, Ltd, 2015: pp. 112–138. 528 [72] C. Pagliero, N. Ochoa, J. Marchese, M. Mattea, Degumming of crude soybean oil by 529 ultrafiltration using polymeric membranes, J Amer Oil Chem Soc. 78 (2001) 793–796. 530 [73] C. Pagliero, M. Mattea, N. Ochoa, J. Marchese, Fouling of polymeric membranes during 531 degumming of crude sunflower and soybean oil, J. Food Eng. 78 (2007) 194–197. 532 [74] J.M.L.N. de Moura, L.A.G. Gonçalves, J.C.C. Petrus, L.A. Viotto, Degumming of vegetable oil 533 by microporous membrane, J. Food Eng. 70 (2005) 473–478. 534 [75] L.G. Peeva, E. Gibbins, S.S. Luthra, L.S. White, R.P. Stateva, A.G. Livingston, Effect of 535 concentration polarisation and osmotic pressure on flux in organic solvent nanofiltration, J. Membr. 536 Sci. 236 (2004) 121–136. 537 [76] I. Sutzkover, D. Hasson, R. Semiat, Simple technique for measuring the concentration 538 polarization level in a reverse osmosis system, Desalination. 131 (2000) 117–127. 539 [77] B. González González, I. Ortiz Uribe, Mathematical Modeling of the Pervaporative Separation 540 of Methanol−Methylterbutyl Ether Mixtures, Ind. Eng. Chem. Res. 40 (2001) 1720–1731. 541 [78] M. Saberi, S.A. Hashemifard, A.A. Dadkhah, Modeling of CO2/CH4 gas mixture permeation 542 and CO2 induced plasticization through an asymmetric cellulose acetate membrane, RSC Adv. 6 543 (2016) 16561–16567. 544 [79] M. Saberi, A.A. Dadkhah, S.A. Hashemifard, Modeling of simultaneous competitive mixed gas 545 permeation and CO2 induced plasticization in glassy polymers, J. Membr. Sci. 499 (2016) 164–171. 546 [80] S. Matsui, D.R. Paul, A simple model for pervaporative transport of binary mixtures through 547 rubbery polymeric membranes, J. Membr. Sci. 235 (2004) 25–30. 548
29
[81] R.S. Prabhakar, R. Raharjo, L.G. Toy, H. Lin, B.D. Freeman, Self-Consistent Model of 549 Concentration and Temperature Dependence of Permeability in Rubbery Polymers, Ind. Eng. Chem. 550 Res. 44 (2005) 1547–1556. 551 [82] D.R. Paul, O.M. Ebra-Lima, Pressure-induced diffusion of organic liquids through highly 552 swollen polymer membranes, J. Appl. Polym. Sci. 14 (1970) 2201–2224. 553 554
Minerva Access is the Institutional Repository of The University of Melbourne
Author/s:
Abdellah, MH; Liu, L; Scholes, CA; Freeman, BD; Kentish, SE
Title:
Organic solvent nanofiltration of binary vegetable oil/terpene mixtures: Experiments and
modelling
Date:
2019-03-01
Citation:
Abdellah, M. H., Liu, L., Scholes, C. A., Freeman, B. D. & Kentish, S. E. (2019). Organic
solvent nanofiltration of binary vegetable oil/terpene mixtures: Experiments and modelling.
